Automatic detection algorithm of intracranial pressure waveform components

We describe an automated detection algorithm that may be used to identify the percussion peak (P), tidal peak (T), dichrotic notch (N), and dichrotic peak (D) components of the intracranial pressure (ICP) signal. The algorithm uses a moving average filter to remove quantization error, a lowpass filter to identify the beat series, and a local search to identify the components of each beat. The algorithm was compared with two experts' visual identification of the percussion components of 997 beats recorded from three subjects. The algorithm accuracy rate was 99.3% with an acceptance interval of 8 ms (/spl plusmn/1 sample).